Metadata and Master Data Management
Metadata describes data’s meaning, structure, lineage, and usage; master data is the shared, authoritative data about core entities. The exam tests metadata types, catalog and lineage use, MDM styles, match/merge, and golden-record governance.
Core rules
- Distinguish business metadata (definitions, owners, policies), technical metadata (schemas, mappings, lineage), and operational metadata (runs, volumes, errors).
- A catalog is only useful if it is connected to lineage, ownership, and quality signals.
- MDM styles: registry (index, source remains authoritative), consolidation (analytical golden record), coexistence (central plus local updates), central (authoritative creation).
- Match and merge requires deterministic and probabilistic rules, thresholds, review queues, and survivorship rules by attribute.
- Golden records need stewards, source precedence, and change history; they are not “the one system.”
Worked example
Three systems hold customer records. The team chooses coexistence MDM: a central hub creates and updates the golden record, while local systems retain operational attributes. Matching uses tax ID plus fuzzy name and address; scores above 95 auto-merge, 80–95 go to review, below 80 stay separate. Survivorship assigns legal name to ERP, preferred name to CRM, and credit status to finance. The reasoning: one style and explicit source precedence prevent both duplicate creation and silent overwrites.
Traps
- Treating metadata as static documentation with no lineage.
- Assuming the golden record replaces every source system.
- Auto-merging without thresholds or review.
- No stewardship, so survivorship rules go stale.